How Do I Improve Computer Performance?

Computer performance refers to the indicators used to measure the performance of computer systems, but the most reliable measure is time. Time can be defined differently according to the calculation method, such as response time, CPU time, etc.

When designing a computer system, in addition to figuring out the hardware interface, from the designer's point of view, the hardware function should be implemented most reasonably according to the best performance and price ratio desired. So what is the performance-price ratio of computer systems? And how to measure them?

There are various scales for measuring computer system performance, but the most reliable measure is time. Time can be defined differently according to the calculation method, such as response time, CPU time, etc. Response time refers to the waiting time after the user sends a task to the computer system until the desired result is obtained. This includes the time to access the disk and access the main memory, the CPU computing time, the I / O action time, and the time overhead of the operating system work. Although this definition is relatively intuitive, for multiple programs, since the CPU can transfer to other programs while waiting for I / O operations, the response time cannot distinguish this situation. The other case is only considering the CPU time, which can be distinguished at this time, it will not include the time waiting for I / O operations and the time spent by the CPU to switch to other programs. Of course, the CPU time itself can be divided into user CPU time and system CPU time. The statistics of the system CPU time are difficult to be accurate, because this actually requires the operating system to perform self-measurement. In addition, when comparing machines with different system codes, since the system CPU time is different, the error is larger, so it is more appropriate to use the user CPU time as a performance measurement time. Of course, when measuring the performance of an unloaded system, it is more appropriate to use response time, and to measure CPU performance is to use user CPU time. The following mainly discusses CPU performance measured by user CPU time

Most computers use a clock that runs at a fixed rate. Its operating cycle is called clocks, and it is often expressed in terms of time or operating rate. The time Tcpu required for a program to run on the CPU can be expressed by the following formula:

Where

Represents the total number of instructions in the program to be executed, CPI represents the average number of clock cycles required to execute each instruction, and Tc represents the length of the clock cycle. As can be seen from the above formula, the user CPU time depends on three characteristics: the clock cycle (or rate), the number of time cycles required for each instruction, and the number of instructions in the program. Which mainly depends on the machine instruction system and compilation technology, CPI is mainly related to computer composition and instruction system, and Tc is mainly determined by hardware technology and computer composition. The average number of clock cycles CPI required by each instruction can be expressed by the following formula:

Among them, Ii indicates the number of executions of the i-type instruction in the program, CPIi indicates the average number of cycles required to execute an i-type instruction, and n is the number of all instruction classes in the program. The above formula can also be rewritten as:

among them,

Represents the proportion of type i instructions in the program ^[1]

MIPS (Million Instruction Per Second) It means million instructions per second. This is a metric used to describe computer performance. For a given program, MIPS can be defined as:

Another alternative standard is MFLOPS (Million Floating Point Operation Per Second), which means millions of floating point operations per second. MFLOPS can be expressed as:

When evaluating the computer system, in addition to the parameters of the structure and function of the machine being evaluated, it is also closely related to the input, that is, the workload of the computer system. A computer system that is evaluated tends to exhibit higher performance for one workload, and may exhibit lower performance for another workload. In order to objectively evaluate the performance of a computer system, it is necessary to select a workload that is truly representative. Different levels of benchmarking procedures are usually used to evaluate system performance.

1. Use actual applications. Compilers such as C or C ++; Photoshop graphics processing software, Premiere and CAD tool software.

2. Use core procedures. This is to extract a small number of key loop program segments from the actual program and use this to evaluate performance, but these core programs have only the value of evaluating performance.

3. Toy benchmarking procedures. Toy benchmarking programs are usually only 10 to 100 lines and the results are predictable. It is popular because it is small and easy to type and works on any machine.

4. Synthesis test program. It is similar to the core program method, but this synthetic test is artificially compiled. Popular synthetic test programs include WinBench 99, 3DMMARK2001, WhatCPUIs, and so on. Especially WinBench 99, the test results provided by it are very convincing and very authoritative. Therefore, when most newspapers introduce new hardware, the data provided are the test results from WinBench 99.

Relevant experts analyzed the above methods and concluded the following deficiencies:

1. The above methods all attempt to use a test data to characterize the processing power of the computer.

2. The above method is only for performance testing, and there is no performance analysis work.

Through the above analysis and proposed a relatively comprehensive and comprehensive performance evaluation method, including two parts:

1. Computer performance test.

2. Analysis of test data.